Drone Tracking Radar: Part 6 Range Doppler

hello again everyone so we've finally made it to the final video of this build your own radar Series so far we've covered radar Hardware we've talked about CW Radar fmcw Radar synchronizing chirps and CFR targeting and now in this video we'll cover multiple chirp fmcw range stopper plotting and then put everything together into a real radar and use it to track a car and a bike and even a small drone okay so so far we've been transmitting these linear frequency modulated chps and we're just doing that over and over and looking at the beat frequency that they produce and that is very interesting but it gets even more interesting if we start keeping track of the chirps so if we number the chirps and keep track of what data we receive from each of them then we'll be able to see how that data is changing from chirp to chirp and so what we do is we take these chirps and we're going to arrange them into columns so we take a trip that we've received and that we've done the max filter on and we separate it out and arrange it as a column and we do that with all of our chirps so for example say we have six chirps so that will give us six Columns of data and let's compress those columns together so we have our six chirps and it gives us this 2D Matrix of data The Matrix is n rows by m columns where n is the number of data points per chirp and M is the number of chirps so that Matrix goes from 0 to n minus one and then from Zer to M minus one so for us m is six and as shown here n is 9 and if you think about it we now have time on the Y AIS but we also have time on the x-axis and so to prevent any confusion I'm going to label the time on the Y AIS I'm going to call that fast time and the time on the xaxis I'll label slow time it's called slow time because if you look across each of the columns those data points are all separated by the pulse repetition interval which is on the of milliseconds but on the Y AIS fast time those data points are all separated by the sampling rate of the data converter so that'll be something like microseconds or less that's why we call them fast time and slow time so hopefully that is clear this is all really important to get straight in your head it basically forms the foundation for all kinds of radar processing okay so now we have this 2D Matrix of our chirp data and we've seen in past videos that if we do an fft on one chirp of data that we will get the range so that is still true we can do an fft on any of these columns and we will get range so none of that has changed so doing that for all of the columns now gives us range on the Y AIS and slow time on the x-axis so again nothing new here we've just organized the chirp such that we're showing the beat frequency which corresponds to the range of each chirp but by putting them together like this we now have a matrix of signal strength across range at regular time intervals and those regular time intervals are of course the pulse repetition rate so if we look across each row we are seeing how that range is changing over time and range changing over time is velocity so to get velocity we take another fft but this time we go across the columns and in radar terms we call that velocity we call that Doppler and we can do both of these ffts in one fill swoop by using a function called the two-dimensional fft this function exists in python or mat lab or whatever and it just generates this transformation very cleanly so for our radar data applying a two-dimensional fft to our fast time slowtime Matrix will return a new Matrix with range on the Y AIS and velocity or Doppler along the x-axis and so the programming for this range Doppler plot is very easy we just separate out the chirps into columns and then apply a 2d fft the timing to separate out those chirps is something we covered in a previous video in this series so if you haven't seen that you can go back and watch that the timing is very important because that timing is what we use to calculate the velocity so each toggle of this gpio burst on the Raspberry Pi is going to kick off a long sequence phase synchronous lfm ramps and it gives us this long buffer of data which is going to contain all of the chps and then we just break that buffer into individual chps and apply the 2D fft to that so let's go ahead and try that now okay so this is what those range Doppler plots look like so this is a video of some street traffic so a car with with a big radar cross-section and a bike with a bit smaller cross-section I'll just let this play for a bit so you can get a feeling for the display of data on these range Doppler plots each frame or update of this range Doppler plot uses 128 fmcw chirps so the Y AIS is range and that goes from 0 m to 40 m and of course 0er meters corresponds to the phaser radar which you can see mounted on that table and then the velocity is just taken from the fft of range across multiple chirps that velocity is plotted on the xaxis in meters per second but I'm just noticing now that the x-axis labels got clipped off uh so sorry about that but that is velocity in meters perss that's that's what it's supposed to say and you can see that the right half of the screen is positive velocity so these would be targets moving away from the radar and the left half of the screen is negative velocity so targets moving towards the radar there's a big line right in the middle of the screen and that corresponds to 0 m/s and we have a name for things moving at 0 m/s and that name is stationary so this is all the stationary items and we call all those things ground clutter we could try to filter this out or sometimes we'll put a notch filter around 0 m/s to just delete all of that and when I show this plot with the drone target I'll put that notch in in so that you can see what that looks like but for this plot I wanted to leave all the ground clutter in there so that you could see what was the raw processing of the range Doppler two-dimensional fft you'll also notice that at a range of 0er MERS that there's a point there that is very bright and that is not due to ground clutter but instead it is due to the transmitting antenna leaking directly into the receiver array so this could be greatly improved by isolating the transmit and receive antennas but on a CW radar system there's always some leakage from transmit to receive and that is because in a CW system both transmit and receive are on at the same time and of course our transmit power is going to be much higher power than anything we could back on the receiver so it's very common to see some kind of bright point at zero velocity and zero range so that is interesting and hopefully it is a good example to explain what range Doppler plotting looks like but I really want to try this radar out on a small drone so let's try that next so this is the Drone that we're going to track it is a very small drone it is the DJI mini3 Pro it is 249 gram and completely made of plastic and of course here's the phaser radar that we'll be using I've attached a pastor neack x-band horn antenna to it I've tried a few different antennas uh like the Vivaldi that comes with the phaser kit as well as some patch antennas and all looked more or less similar at least for the stuff that I'm doing but certainly this horn antenna is the best for single Channel transmit but before we try radar on it I also want to show you what our transmitter waveform and power output look like this is a field Fox Spectrum analyzer from kyite a spectrum analyzer plots the RF power versus frequency so basically the same as what we were doing when we took our ffts although this is a vastly better implementation than what we can do with Pluto and this spectrum analyzer can plot that power over a frequency range from nearly DC all the way up to 26 GHz so that's quite a bit more frequency range than what our f laser and Pluto can do and that frequency range allows us to see the complete picture of transmit power as well as transmit frequency and any other Spurs that we are transmitting on and you can see that there is only one frequency Spike on this plot which is great that is our transmit frequency of 10.3 GHz there are no other Spurs that Peak above the noise floor of minus 60 dbm so that's great that is exactly what we want but you can also clearly see that the transmit power is only about minus 3 dbm so that is is 1/ half of 1 M that is very low output power for a radar a normal CW radar for drone tracking might be something like 10 to 50 Watts so that's tens of thousands of times more output power now we could easily increase the output power of phaser by adding an amplifier uh and maybe I'll do that in a future video but really the goal of these videos is to show you how to do this or at least the basics on how this works and how to get started so I'll probably just leave the improvements and Perfections up to you in your project project there is always so much more that you could do uh both on the harbor side and on the sofware side okay so let's try phaser and see how well it can spot a small drone so here's that small DJI drone flying around in my backyard the total depth of my yard is about 20 M and as the Drone moves back and forth you can see it as that small bright spot you can also see its propeller signature that is called micro Doppler this is due to the propeller blade spinning so even when the Drone is stationary you can still easily distinguish it from the ground clutter by those spinning propeller blades and the micro doopler signature turns out to be important data that people use to distinguish and classify drones we also saw micro Doppler with the car and bike video and that was from the wheels rotating so feel free to rewind back to that and see if you can spot the micro doler in that video but it wasn't quite as bright or noticeable as these drone propellers okay so hopefully that was cool and a good look at how to use this range stop PL plotting one final thing before I close out this video and that is this thing called the radar data Cube you may have heard of it before or run across it elsewhere is a very common thing that radar people like to talk about so what is it well right now we have a radar data plane it's fast time and slow time for one antenna element but we could add another axis and that axis is the number of receivers and each new receiver brings with it a new data plane so here's what a second receiver would look like and we can add a third receiver and of course we can keep adding antenna elements and so now it's kind of starting to look like a cube technically the shape is a rectangular cuboid but everybody just calls it the radar data Cube and with the data Cube you can slice it up in interesting ways so here's a slice that is just one element and here's a slice that is all of the elements in one chirp and here's a slice that is all of the elements in one range pin and I should point out that the cube rotates based on what you want to emphasize which means that unfortunately there is not a consistent way to look at representations of the cube any of the sides could be chirps or time samples or antenna elements but you can see that each of these data cubes has all three of those things the number of chirps the number of samples per chps and the number of receivers actually the radar data cube is a next segue into what I hope will be my next video series and that series will be on phased aray beam forming Radars and here's a little preview of what that is going to look like so this is a true two-dimensional scan while operating the radar and it takes advantage of the phased aray ability to electronically steer the antenna so in this next video series I'll show you how to do this and we'll use all the receive elements on phaser to learn some beamforming fundamentals and then we'll apply those fundamentals to this radar system so with that we can do true two-dimensional plots of range and cross range we can do star Imaging we can do adaptive cancellation algorithms just all kinds of cool stuff but that is all going to be in the next series so I hope you enjoyed this series and thank you again for watching

Channel: Jon Kraft

Views: 35,062

Keywords:

Id: xnaaD9Um3K8

Length: 12min 14sec (734 seconds)

Published: Tue Jun 11 2024